Skip to content

88plug/ooda

Repository files navigation

OODA

Route any decision or action through John Boyd's real OODA loop for Claude Code + Grok — Observe, Orient (the schwerpunkt), Decide, Act, loop fast. Not a 4-box speed circle.

plugin-validate License: FSL-1.1-ALv2 Docs Claude Code plugin DeepWiki

OODA (Col. John Boyd) is a theory of winning through faster, better adaptation — not a checklist and not a race. This plugin runs the real loop at the moment you decide, act, iterate, diagnose, or compete.

The anti-patterns that void it, up front: it is not a four-box speed circle, it is not a pure speed race, and Orient is not one step of four — it dominates. The killer failure is Orient-blindness: acting fast on an un-reframed, biased model.

Install

Claude Code

/plugin marketplace add 88plug/claude-code-plugins
/plugin install ooda@88plug

Grok Build

grok plugin marketplace add 88plug/claude-code-plugins
grok plugin install ooda@88plug --trust

The hooks and the bundled skill load automatically from the manifest. No MCP server, no database, no dependencies beyond python3 (routed through a thin-PATH-safe resolver). install.sh is a no-op.

What it does

Hook Event What it does
prime.sh SessionStart Sets the OODA posture once per session
nudge.sh UserPromptSubmit Fires only when the prompt is a decision/action/iteration/incident/compete turn; silent on mechanical requests

Detection lives in hooks/lib/detect.py with a per-session anti-nag cooldown so the same nudge does not fire on back-to-back turns. /ooda:loop runs the loop on demand.

The loop (real, not the cartoon)

Phase Boyd's meaning Do
Observe intake unfolding circumstances + your interaction with them measure, don't assume — your prior acts changed the field
Orient the schwerpunkt: genetic/cultural/experience priors + new info + analysis & synthesis reframe against stale models (the highest-leverage step)
Decide "Decision (Hypothesis)" pick one lever, as a hypothesis
Act "Action (Test)" ship a test; the result re-enters Observe

Two implicit-guidance arrows run Orient→Observe and Orient→Act, bypassing explicit Decide — the trained-operator fast path for proven moves.

Orient dominates

Orientation shapes what you observe and how you decide and act, and is reshaped by feedback. On a surprise or a failing baseline it runs destruction and creation: break the stale frame into its parts, synthesize a new hypothesis — because Gödel, Heisenberg, and the 2nd law guarantee any closed model drifts from reality. Orientation is never finished.

Tempo is not speed

Getting "inside the loop" is unpredictable rhythm and variety that outpaces the problem's Orient — not stopwatch haste. A predictable fast loop is counterable. For work: cheapest-falsifier-first, short varied revert-safe cycles that re-orient before your own model goes stale.

Composes with

  • cynefin — Cynefin lives inside Orient: classify the domain by its constraints before choosing the response.
  • scientific-method — Observe and Act are the experiment: the cheapest-falsifier probe and the ship-as-a-test.
  • break-dogma — a stale previous-experience input to Orient is exactly the inherited assumption to test.

Test

bash tests/smoke.sh                # prime injects; nudge fires on decision/action language, silent on mechanical; cooldown suppresses a repeat
python3 .ci/validate_plugin.py .   # manifest, keywords, hook paths, skill frontmatter

Credits

The framework is John Boyd's. This plugin packages his published work (the briefings collected in A Discourse on Winning and Losing; Frans Osinga's Science, Strategy and War; Chet Richards; Robert Coram) as a routing discipline — see the skill's Sources section.

License

FSL-1.1-ALv2.

Releases

Packages

Contributors

Languages